[squeak-dev] The Trunk: Monticello-ct.715.mcz

Thiede, Christoph Christoph.Thiede at student.hpi.uni-potsdam.de
Sat Apr 11 14:42:02 UTC 2020


Hi Jakob, hi all,


> Did you intentionally not reply the mailing list?


I could not find your previous mail on the list, too! However, we're back on the list now :)

Hm ... what about deprecating #orderedItems and introducing a new #allItems instead? Then we could avoid any internal confusion and had a stable public protocol.

> Which reminds me in general that Monticello classes could use some class comments. :-(

A really good idea! :D

Best,
Christoph
________________________________
Von: Jakob Reschke <forums.jakob at resfarm.de>
Gesendet: Samstag, 11. April 2020 15:53:11
An: Thiede, Christoph
Betreff: Re: [squeak-dev] The Trunk: Monticello-ct.715.mcz

Hi Christoph,

Did you intentionally not reply the mailing list?

We should care, the method is public and AFAIK the class is also not in some private category. Also "outsiders" should not care about the instance variable, but about the interface. Which was altered by this change.

Which reminds me in general that Monticello classes could use some class comments. :-(

Kinds regards,
Jakob


Thiede, Christoph <Christoph.Thiede at student.hpi.uni-potsdam.de<mailto:Christoph.Thiede at student.hpi.uni-potsdam.de>> schrieb am Sa., 11. Apr. 2020, 14:41:

Hi Jakob, sorry for the late reply!


My doubts about option 2 would be that as a developer of the sorter, you might not expect "self orderedItems" to return a different result than "orderedItems" (the instance variable). To me, this feels a bit like an anti-pattern.


I don't know whether we have to care about compatibility outside the Trunk: If not, we could introduce a new getter such as #allOrderedItems for this purpose ...


Best,

Christoph

<http://www.hpi.de/>
________________________________
Von: Jakob Reschke <forums.jakob at resfarm.de<mailto:forums.jakob at resfarm.de>>
Gesendet: Freitag, 10. April 2020 08:39:31
An: Thiede, Christoph
Betreff: Re: [squeak-dev] The Trunk: Monticello-ct.715.mcz

Hi,

Is there a use case for not getting all the items, or getting an
intermediate list? There wasn't before, right? I would find it strange
if I had to tell the sorter "give me what you have got, I mean all of
it". This would be unusual and presumably easy to forget. So without
further arguments, I would pick option 2.

Kind regards,
Jakob

Am Fr., 10. Apr. 2020 um 00:33 Uhr schrieb Thiede, Christoph
<Christoph.Thiede at student.hpi.uni-potsdam.de<mailto:Christoph.Thiede at student.hpi.uni-potsdam.de>>:
>
> Hi Jakob,
>
>
> Thanks for the report - and I'm sorry to hear that I broke this!
>
>
> The motivation for the deferred list was that #addToOrder: (and so the constructor #items:) actually does not sort the added items, but only their dependencies. That's why you need to send #addDeferred to the sorter after adding all other items. Before now, I did not realize that #orderedItems has external users.
>
>
> So there are two options:
>
> 1. Permanently change the public protocol of the dependency sorter - if you want to get all items, you need to send #addDeferred ti the sorter before retrieving the #orderedItems. In the Trunk, we would need to update four senders in three classes for this solution.
>
> 2. Change #orderedItems to lazily append the deferred items to the returned list. This is not ideal as it makes the getter more expensive, but on the other hand, we can preserve compatibility by doing so.
>
>
> I'm open for your opinion on both approaches, or for any other idea! :-)
>
>
> Best,
>
> Christoph
>
> ________________________________
> Von: Jakob Reschke <forums.jakob at resfarm.de<mailto:forums.jakob at resfarm.de>>
> Gesendet: Freitag, 10. April 2020 00:03:47
> An: Thiede, Christoph; squeak-dev at lists.squeakfoundation.org<mailto:squeak-dev at lists.squeakfoundation.org>
> Betreff: Re: [squeak-dev] The Trunk: Monticello-ct.715.mcz
>
> Forgot to mention the net result of the issue: the postscripts gets
> effectively removed in the saved version.
>
> Am Fr., 10. Apr. 2020 um 00:02 Uhr schrieb Jakob Reschke
> <forums.jakob at resfarm.de<mailto:forums.jakob at resfarm.de>>:
> >
> > It looks like this broke the writing of postscripts in FileTree repositories.
> >
> > MCFileTreePackageStructureStWriter>>acceptVisitor:forDefinitions:
> > (MCDependencySorter sortItems: aCollection) does not include an
> > MCPostscriptDefinition that is still present in aCollection.
> >
> > MCDependencySorter class>>sortItems:
> > The sorter includes the MCPostscriptDefinition in the new deferred
> > list, but not in the orderedItems. Therefore it is not included in the
> > result of sortItems:, because it only returns the orderedItems, not
> > the deferred ones.
> >
> > Christoph, can and must sortItems: be fixed or a different method?
> >
> >
> > Am Do., 5. März 2020 um 21:05 Uhr schrieb <commits at source.squeak.org<mailto:commits at source.squeak.org>>:
> > >
> > > Nicolas Cellier uploaded a new version of Monticello to project The Trunk:
> > > http://source.squeak.org/trunk/Monticello-ct.715.mcz
> > >
> > > ==================== Summary ====================
> > >
> > > Name: Monticello-ct.715
> > > Author: ct
> > > Time: 6 February 2020, 6:58:26.533648 pm
> > > UUID: 199bf8a3-5eee-1045-a44b-89ad30fa5523
> > > Ancestors: Monticello-cmm.708
> > >
> > > Fixes wrong load and unload order of MCScriptDefinitions
> > >
> > > See http://forum.world.st/Monticello-Bug-Preamble-of-removal-is-executed-too-late-tp5108401.html. Regression tests are in Tests-ct.426.
> > >
> > > Please review!
> > >
> > > =============== Diff against Monticello-cmm.708 ===============
> > >
> > > Item was added:
> > > + ----- Method: MCDefinition>>wantsToBeOutermost (in category 'testing') -----
> > > + wantsToBeOutermost
> > > +
> > > +       ^ false!
> > >
> > > Item was changed:
> > >   Object subclass: #MCDependencySorter
> > > +       instanceVariableNames: 'required provided deferred orderedItems'
> > > -       instanceVariableNames: 'required provided orderedItems'
> > >         classVariableNames: ''
> > >         poolDictionaries: ''
> > >         category: 'Monticello-Loading'!
> > >
> > > Item was changed:
> > >   ----- Method: MCDependencySorter>>add: (in category 'building') -----
> > >   add: anItem
> > >         | requirements |
> > > +       (anItem wantsToBeOutermost)
> > > +               ifTrue: [^self addDeferred: anItem].
> > >         requirements := self unresolvedRequirementsFor: anItem.
> > >         requirements isEmpty
> > >                 ifTrue: [self addToOrder: anItem]
> > >                 ifFalse: [self addRequirements: requirements for: anItem].
> > >         ^anItem!
> > >
> > > Item was added:
> > > + ----- Method: MCDependencySorter>>addDeferred (in category 'building') -----
> > > + addDeferred
> > > +       deferred do: [:ea |
> > > +               self addToOrder: ea].!
> > >
> > > Item was added:
> > > + ----- Method: MCDependencySorter>>addDeferred: (in category 'private') -----
> > > + addDeferred: anItem
> > > +       ^ deferred add: anItem!
> > >
> > > Item was changed:
> > >   ----- Method: MCDependencySorter>>initialize (in category 'initialize-release') -----
> > >   initialize
> > >         provided := Set new.
> > >         required := Dictionary new.
> > > +       orderedItems := OrderedCollection new.
> > > +       deferred := OrderedCollection new.!
> > > -       orderedItems := OrderedCollection new.!
> > >
> > > Item was changed:
> > >   ----- Method: MCPackageLoader>>sorterForItems: (in category 'private') -----
> > >   sorterForItems: aCollection
> > >         | sorter |
> > >         sorter := MCDependencySorter items: aCollection.
> > >         sorter addExternalProvisions: self provisions.
> > > +       sorter addDeferred.
> > >         ^ sorter!
> > >
> > > Item was added:
> > > + ----- Method: MCPostscriptDefinition>>wantsToBeOutermost (in category 'nil') -----
> > > + wantsToBeOutermost
> > > +
> > > +       ^ true!
> > >
> > > Item was added:
> > > + ----- Method: MCRemovalPreambleDefinition>>wantsToBeOutermost (in category 'testing') -----
> > > + wantsToBeOutermost
> > > +
> > > +       ^ true!
> > >
> > >
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20200411/0fc642c3/attachment.html>


More information about the Squeak-dev mailing list